The technique of subtitling for Audio-Visual (AV) material has been used beginning with the first celluloid cinema movies and further until the recent digital media appeared. The main target of subtitling has been the support of handicapped people or small ethnographic language groups. Therefore subtitling often aims at the presentation of text information even when having been encoded as graphic data like pixel maps. Therefore pre-produced AV material for broadcasting (Closed Caption, Teletext, DVB-Subtitle etc.) and movie discs (DVD Sub-Picture etc.) primarily are optimized for subtitles representing simple static textual information. However, progress in PC software development for presentation and animation of textual information induces a corresponding demand for possibilities and features within the digital subtitling technique used for pre-recording and broadcasting. Using straightforward approaches without any special precautions, these increased requirements for subtitling would consume a too big portion of the limited overall bandwidth. The conflicting requirements for a ‘full feature’ subtitle encompassing karaoke all through genuine animations are on one hand the coding efficiency and on the other hand the full control for any subtitle author.
For today's state of the art of digitally subtitling AV material with separate subtitling information two main approaches exist: Subtitling can be based on either pixel data or on character data. In both cases, subtitling schemes comprise a general framework, which for instance deals with the synchronization of subtitling elements along the AV time axis.
Character Data Based Subtitling:
In the character-based subtitling approach, e.g. in the teletext system ETS 300 706 of European analog or digital TV, strings are described by sequences of letter codes, e.g. ASCII or UNICODE, which intrinsically allows for a very efficient encoding. But from character strings alone, subtitling cannot be converted into a graphical representation to be overlaid over video. For this, the intended character set, font and some font parameters, most notably the font size, must either be coded explicitly within the subtitling bitstream or an implicit assumption must be made about them within a suitably defined subtitling context. Also, any subtitling in this approach is confined to what can be expressed with the letters and symbols of the specific font(s) in use. The DVB Subtitling specification ETS 300 743, in its mode of “character objects”, constitutes another state-of-the-art example of character-based subtitling.
Pixel Data Based Subtitling:
In the pixel-based subtitling approach, subtitling frames are conveyed directly in the form of graphical representations by describing them as (typically rectangular) regions of pixel values on the AV screen. Whenever anything is meant to be visible in the subtitling plane superimposed onto video, its pixel values must be encoded and provided in the subtitling bitstream, together with appropriate synchronization info, and hence for the full feature animation of subtitles all pixel changed must be transported. Obviously, when removing any limitations inherent with full feature animations of teletext, the pixel-based approach carries the penalty of a considerably increased bandwidth for the subtitling data. Examples of pixel-based subtitling schemes can be found in DVD's sub-picture concept “DVD Specification for Read-Only disc”, Part 3: Video, as well as in the “pixel object” concept of DVB Subtitling, specified in ETS 300 743.